【作者】 王燕花；

【导师】 简水生；

【作者基本信息】 北京交通大学 ， 通信与信息系统， 2009， 博士

【摘要】 光纤传感正在向着网络化、智能化的方向发展。本论文结合北京交通大学科技专项基金资助项目“光纤传感预警系统的研究”,以实现实用化的大型智能光纤传感系统为目的,针对相干FBG组传感系统、MZI光纤传感系统以及光纤传感信号的自动识别等进行了深入的理论和实验研究,获得的创新性成果如下:·深入分析了相干FBG组传感系统的性能。采用改进的耦合模理论及Rouard算法,分析了光栅初始相位对相干FBG组透射谱的影响;推导得到估算相干均匀及啁啾FBG组FSR的简单公式,与复杂的数值仿真结果有较好的一致性;推导得到相干FBG组测量时变信号的频率极限条件;采用增大距离的方式减敏,准确解调振动信号频率,在100Hz至15000Hz频段测得频率与压电陶瓷传感器一致,高频响应性能优于压电陶瓷传感器;从相位波动及频率分布角度研究激光器线宽对相干FBG组传感灵敏度的影响,理论和实验分析得到当激光器全高半宽FWHM小于相干FBG组FSR的1%时,可以获得较好的传感效果。·构建了3×3耦合器解调与互相关定位相结合的双向MZl分布式光纤传感系统,对其传感及定位性能进行了深入分析。对3×3耦合器非理想特性及采集卡采样速率和非同步性可能引入的解调误差进行了深入分析;通过观察支路波形单周期极值点个数,采用波动范围较大的信号进行预校正,降低了解调误差;提出用于估计干涉信号最高频率的经验公式;结合双向MZI传感系统结构特点,将互相关运算限制在较短的时间序列范围内,减少了计算量;通过对比实验,采用滑动均值滤波降噪,并对互相关函数进行归一化、周期化及最小化处理,提高了定位准确性。·实现了光纤传感信号的自动识别。对相干FBG组及光纤MZI传感器的具体结构进行了设计,分别针对微弱振动信号及强干扰信号,采用缠绕法、平面法、特殊材料涂覆光纤及弹簧隔离法、整体隔离法等设计光纤传感器,提高了对目标信号的识别能力;通过提取信号最大幅度、有效电平、振铃计数、脉冲计数、事件计数、持续时间等参数,对信号进行傅立叶谱分析、AR谱分析、时频分析,并对不同参数及频谱分析结果再进行分布统计分析,实现了多角度的信号物理特征提取,应用于列车区间定位、列车计轴、围栏破坏性行为监测、管道破坏性行为监测等场合,实现了各种不同目标信号的特征提取;采用基于Fisher准则的线性判别函数,实现了基于相干FBG组传感系统的列车区间空闲或占用的自动识别;采用LMBP人工神经网络,实现了基于3×3耦合器解调的MZI传感系统的围栏破坏性行为的自动识别。更多还原

【Abstract】 Optical fiber sensing system is becoming large-scale and intelligent.To realize practical large scale intelligent optical fiber sensing system,this dissertation focus on coherent FBG group sensing system,MZI fiber sensing system and auto-recognition of fiber sensing signals.The work is supported by the Foundation of Beijing Jiaotong University(2006XZ010).And the main achievements of the dissertation are listed as follows:●The performances of coherent FBG group sensing system are investigated. Based on the improved coupled-mode theory and Rouard method,the influence of the initial phase of FBG on the spectrum of coherent FBG group is proved.Simplified formulas are derived for the estimation of FSR for coherent FBG group formed by both uniform FBG and chirped FBG,and the estimation results are in good agreement with those calculated by more complicated numerical methods.The frequency limitation for measurement of time-varying signals by coherent FBG group is studied. Desensiblization is implemented by increasing the distance and the frequency of the vibration signal is demodulated more accurately.The frequency measurement results of coherent FBG group from 100Hz to 150000Hz consist with those of PZT sensor,and the sensing performance of coherent FBG group at high frequency band is even better than that of PZT sensor.The influence of the line width of laser on the sensitivity of coherent FBG group is studied from both phase fluctuation and frequency distribution. Theoretical and experimental analysis both prove that a better sensing performance is achieved when the ratio of laser FWHM and coherent FBG group FSR is smaller than 1%.●A bidirectional MZI distributed fiber sensing system with demodulation by 3×3 coupler and location by cross correlation is constructed,and both its sensing and location performances are investigated.The possible errors induced by the imperfect 3×3 coupler and by the improper sampling rate and nonsynchronous of the DAQ device are studied in detail.A method of initial calibration by signals that fluctuate strongly is proposed,and the calibration signal is selected by the number of extreme points of a single period of each branch,by which the demodulation error is greatly reduced.An empirical formula is proposed to estimate the highest frequency of interference signal. Based on the structure of the bidirectional MZI sensor system,the cross-correlation is limited within a short time series and the calculation amount is decreased.The smooth filter is adopted for noise reduction based on contrast experiments,and the cross-correlation is normalized,periodized and minimized,all of which increase the accuracy of the location.●The auto-recognition of fiber sensing signals is realized.Special structures for coherent FBG group and fiber MZI sensors are designed.For weak signals or strongly interferenced signals,winding method,flat layout method,special coated fiber method or spring separation method,solid separation method are proposed,and the recognition ability for object signals is improved.The physical features of the object signals are extracted from various aspects by the parameters analysis such as maximum amplitude, RMS,ringing count,pulse count,event count and duration,by the waveform transform analysis such as Fourier analysis,AR spectrum analysis and joint-time-frequency analysis,and by statistic analysis of the parameters analysis results or waveform transform analysis results.And the features of different object signals are extracted when the optical fiber sensing systems are applied for railway location,axel counter, fence destruction monitoring and pipe destruction monitoring.Auto-recognition of the railway location signals with coherent FBG sensing system is realized by a linear discrimination function based on Fisher criterion.And auto-recognition of the fence destruction signals with a MZI sensing system demodulated by a 3×3 coupler is realized by a LMBP neural network.更多还原